Hybrid Face Recognition System Based on Linear Discriminant Analysis and Voting
Jamal Ahmad Dargham, Ali Chekima, Ervin Gubin Moung, Munira Hamdan · International journal of imaging and robotics · 2014
In this paper, a component-based face recognition system is proposed. Direct matching was used for small component size and Linear Discriminant Analysis (LDA) for larger component size. The outputs from the individual component matching are then combined to achieve the final recognition result. The effect of component size and the image size on the performance of the system are investigated. It is found that direct matching outperforms LDA for small image size while the reverse is true for large image size. In addition, larger component size performs better when applied with LDA while smaller component size can be matched better by using direct matching. It was also found that image size ratio of 1.2 gives better performance rate rather than using square sized image using direct matching.